Name
Title | ||
|---|---|---|
Similarity-Based Heterogeneous Graph Attention Network For Knowledge-Enhanced Recommendation |
Abstract | ||
|---|---|---|
The introduction of knowledge graphs (KG) has improved the accuracy and interpretability of recommendations. However, the performance of KG-based recommender system is still limited due to the lack of valid modeling of user/item similarity and effective constraints on user/item embeddings learning. In addition, common sampling and propagation methods for homogeneous graphs do not apply to KGs due to their heterogeneity. In this work, we propose Similarity-based Heterogeneous Graph Attention Network (SHGAT), which learns both the collaborative similarity and knowledge similarity of items by pre-training item embeddings with user-item interaction data and knowledge propagation in the KG. Meanwhile, users are represented by the items they have interacted with, thus establishing similarity between users and strengthening the learning of item embeddings. Besides, we design an importance sampling and aggregation method based on attention mechanism for heterogeneous graphs. We apply the proposed model on two real-world datasets, and the empirical results demonstrate that SHGAT significantly outperforms several compelling state-of-the-art baselines. |
Year | DOI | Venue |
|---|---|---|
2021 | 10.1007/978-3-030-82147-0_40 | KNOWLEDGE SCIENCE, ENGINEERING AND MANAGEMENT, KSEM 2021, PT II |
Keywords | DocType | Volume |
Recommender systems, Knowledge graph, Similarity modeling, Heterogeneous graph attention network, Importance sampling | Conference | 12816 |
ISSN | Citations | PageRank |
0302-9743 | 0 | 0.34 |
References | Authors | |
0 | 4 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Fan Zhang | 1 | 0 | 9.13 |
| Rui Li | 2 | 0 | 0.34 |
| Ke Xu | 3 | 0 | 0.34 |
| Hongguang Xu | 4 | 0 | 0.34 |